Three-dimensional pattern recognition method to detect shapes in medical images

Abstract

A detection and classification method of a shape in a medical image is provided. It is based on generating a plurality of 2-D sections through a 3-D volume in the medical image. In general, there are two steps. The first step is feature estimation to generate shape signatures for candidate volumes containing candidate shapes. The feature estimation method computes descriptors of objects or of their images. The second general step involves classification of these shape signatures for diagnosis. A classifier contains, builds and / or trains a database of descriptors for previously seen shapes, and then maps descriptors of novel images to categories corresponding to previously seen shapes or classes of shapes.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS[0001]This application is cross-referenced to and claims priority from U.S. Provisional Application 60 / 415,269 filed Sep. 30, 2002, which is hereby incorporated by reference.RESEARCH OR DEVELOPMENT[0002]The present invention was supported in part by grant number RO1 CA72023 from the National Institutes of Health (NIH / NCI). The U.S. Government has certain rights in the invention.FIELD OF THE INVENTION[0003]The present invention relates generally to medical imaging. More particularly, the present invention relates to methods for differentiating normal and abnormal anatomical shapes through a feature estimation method and a classifier.BACKGROUND[0004]Colon cancer is the second leading cause of cancer deaths in the United States. American adults have 1 / 20 chance of developing and 1 / 40 chance of dying from this disease. There are approximately 150,000 new cases diagnosed each year resulting in 56,000 deaths (See e.g. P J Wingo (1995) in a paper enti...

AI Technical Summary

Benefits of technology

[0015]The advantage of the present invention is that it does not make any assumptions about the to-be-detected shape as is common in prior art methods. Another advantage is that if a pre-processing method is used, and method of present invention is used as a post-processing method, then the present method does not have to make assumptions of the pre-processing method. Yet another advantage is that it eliminates the guesswork for obtaining the distinguishing features.

[0016]The objectives and advantages of the present invention will be understood by reading the following detailed description in conjunction with the drawings, in which:

[0017]FIG. 1 shows examples of polyps (a-c) and of healthy tissue (d-e) that have similar shapes;

[0018]FIGS. 2-3 show a general overview of the method according to the present invention;

[0019]FIG. 4 shows an overview of the ROSS method according to the present invention;

[0020]FIG. 5 shows an example of three cross sections of a body through the anatomical directions (a) axial, (b) coronal, (c) sagittal, shows an example of two 3-D renderings of a polyp with different camera views (d), and shows random oriented, mutually orthogonal triple planes through this polyp (e-g), all according to the present invention;

Problems solved by technology

Unfortunately, colon cancer is most often discovered after the patient develops symptoms, and by then, the likelihood of a cure has diminished substantially.

However, FOC is not feasible as a population screening test due to cost, the small but real risk of complications such as perforation, and due to the fact that there are not sufficient endoscopists in the country to accommodate all patients.

The initial clinical results are quite promising, yet the technique is still impractical due, in part, to the time required to review hundreds of images per patient study.

This limitation begs for a computer-aided detection (CAD) method to help the radiologist detect polyps efficiently from the acquired CTC data.

Identifying colonic polyps using CAD is challenging because they come in various sizes and shapes, and because thickened folds and retained stool may mimic their shape and density. FIG. 1 demonstrates the appearance of polyps and healthy tissue as they appear in a virtual colonoscopy study.

Some of the methods incorporate simple intuitions about the shapes of polyps and non-polyps, which leads to false positive detections.

However, polyps span a large variety of shapes, and fitting spheres alone is not an accurate measure.

As mentioned supra, polyp recognition is a difficult problem, but so is the manual identification of discriminating criteria between polyps and healthy tissue.

Such an examination is costly, time consuming and inefficient.

Images